$ oc edit network.operator cluster --subresource=status
Monitor network traffic immediately after cluster creation by automatically installing network observability during the installation process. This Technology Preview feature requires enabling the NetworkObservabilityInstall feature gate.
|
{FeatureName} is a Technology Preview feature only. Technology Preview features are not supported with Red Hat production service level agreements (SLAs) and might not be functionally complete. Red Hat does not recommend using them in production. These features provide early access to upcoming product features, enabling customers to test functionality and provide feedback during the development process. For more information about the support scope of Red Hat Technology Preview features, see Technology Preview Features Support Scope. |
Technology Preview clusters with the NetworkObservabilityInstall feature gate enabled automatically install the Network Observability Operator during cluster creation.
|
Automatic network observability installation is a Technology Preview feature only. Technology Preview features are not supported with Red Hat production service level agreements (SLAs) and might not be functionally complete. Red Hat does not recommend using them in production. These features provide early access to upcoming product features, enabling customers to test functionality and provide feedback during the development process. For more information about the support scope of Red Hat Technology Preview features, see Technology Preview Features Support Scope. |
|
The |
On OKD 5.0 clusters with the NetworkObservabilityInstall feature gate enabled, the Network Observability Operator and its dependencies are automatically installed and configured during cluster creation. This provides immediate access to network traffic monitoring capabilities through the OKD web console.
You can use network observability to complete the following tasks:
Monitor network traffic between pods, services, and external endpoints.
Identify network performance issues such as packet loss and latency.
Troubleshoot connectivity problems between applications.
Analyze network usage patterns for capacity planning.
Detect security anomalies in network behavior.
Network Observability automatic installation is a Technology Preview feature that requires the NetworkObservabilityInstall feature gate. To use this feature during cluster installation, you must complete the following tasks:
Enable the NetworkObservabilityInstall feature gate by creating a manifest file during cluster installation.
Accept that enabling the CustomNoUpgrade feature set prevents minor version updates.
Understand that clusters installed with this Technology Preview feature cannot be upgraded when the feature reaches general availability (GA). You must create a new cluster using the GA version.
When the NetworkObservabilityInstall feature gate is enabled, the following cluster types enable network observability by default during cluster installation:
Standard multi-node clusters
Hosted Control Plane clusters
The following cluster types do not enable network observability by default:
Single-node OpenShift clusters, due to resource constraints. You can override this by setting installationPolicy to InstallAndEnable in the installation configuration.
Clusters where you set installationPolicy to NoAction in the installation configuration.
After cluster installation completes, you can access network observability features immediately without additional configuration:
View cluster-wide network traffic metrics and identify top talkers from Observe → Network Traffic in the web console.
Explore visual representations of network traffic flows between namespaces, workloads, and nodes.
Access Prometheus metrics and Grafana dashboards that track network performance indicators including throughput, packet drops, and connection counts.
The default network observability configuration balances monitoring capabilities with cluster resource usage. On typical clusters, network observability consumes approximately 3-5% of CPU and memory resources per node.
The automatic installation creates a FlowCollector resource with the following default configuration:
Agent type: eBPF with DNS tracking enabled
Sampling interval: 400 (collects 1 in every 400 flows)
Deployment model: Service
Loki storage: None (metrics are stored, but flows are not persisted)
Namespace: netobserv
|
To enable flow persistence and features that require flow records, you must configure a |
If you need to reduce resource consumption, you can customize the FlowCollector resource configuration to adjust sampling rates, filter traffic, or disable specific features.
The automatic network observability installation occurs during cluster creation in OKD 5.0 or later, or when upgrading from OKD 4.22 to 5.0. The Cluster Network Operator (CNO) installs the Network Observability Operator and creates a default FlowCollector custom resource (CR).
The installation process typically completes within a few minutes after the cluster becomes available. If installation fails, the CNO automatically rolls back the changes.
The CNO does not monitor or manage the Network Observability Operator after installation attempts complete. If you delete the Network Observability Operator or the FlowCollector resource, they are not automatically reinstalled.
The CNO does not manage the FlowCollector resource because it is managed by the Network Observability Operator.
|
To reset the installation state and allow the CNO to retry automatic installation, remove the
Delete the condition entry where |
To remove network observability, see "Uninstalling Network Observability Operator".
|
To prevent automatic installation during an upgrade from OKD 4.22 to 5.0, set |
The cluster installation process automatically deploys network observability capabilities when you run the openshift-install program with the NetworkObservabilityInstall feature gate enabled. You do not need to add network observability configuration to your install-config.yaml file unless you want to override the default behavior.
If the NetworkObservabilityInstall feature gate is not enabled, automatic network observability installation does not occur during cluster creation.
When the feature gate is enabled, you can choose one of the following outcomes:
Standard multi-node and hosted control planes clusters enable network observability by default. You do not need to add or change the networkObservability section in your install-config.yaml file.
With this default configuration:
The Cluster Network Operator (CNO) automatically installs network observability on standard multi-node clusters and hosted control planes clusters.
Single-node OpenShift clusters do not install network observability by default due to resource constraints.
Add the following configuration to your install-config.yaml file to prevent installation on all cluster types:
networking:
networkObservability:
installationPolicy: NoAction
Add the following configuration to your install-config.yaml file to install network observability on all cluster types, including single-node OpenShift clusters:
networking:
networkObservability:
installationPolicy: InstallAndEnable
During cluster installation, the following sequence occurs:
You optionally configure the networking.networkObservability.installationPolicy field in the install-config.yaml file to override the default behavior.
The openshift-install program generates the Network CR using the network observability configuration in the install-config.yaml file.
The Cluster Network Operator (CNO) reads the installationPolicy field in the Network custom resource (CR) if it exists. Based on the installationPolicy value and your cluster type, the CNO applies the default, NoAction, or InstallAndEnable outcome.
If the policy enables installation on your cluster type, the CNO:
Installs the Network Observability Operator
Creates a default FlowCollector configuration that starts collecting network flow data
If the effective policy is NoAction, the CNO skips network observability installation.
|
The automatic installation process deploys the Network Observability Operator in the |
When you override the default behavior, the install-config.yaml file contains the network observability configuration. The following example sets installationPolicy to InstallAndEnable:
apiVersion: v1
baseDomain: example.com
networking:
networkObservability:
installationPolicy: InstallAndEnable
The Network custom resource (CR) copies this configuration during cluster creation:
apiVersion: config.openshift.io/v1
kind: Network
metadata:
name: cluster
spec:
networkObservability:
installationPolicy: InstallAndEnable
The CNO installs the Network Observability Operator only once during cluster creation or upgrade. After installation, the CNO does not monitor or reinstall the Operator if you remove it.
The CNO supports only two installation policies: InstallAndEnable and NoAction. The CNO never removes network observability components.
|
To prevent automatic installation during an upgrade from OKD 4.22 to 5.0, set |
When the NetworkObservabilityInstall feature gate is enabled, disable automatic installation of the Network Observability Operator during cluster creation by setting installationPolicy to NoAction in your install-config.yaml file.
|
Since automatic network observability installation is a Technology Preview feature in OKD 5.0, network observability is not installed automatically unless you enable the |
When the NetworkObservabilityInstall feature gate is enabled, OKD installs the Network Observability Operator automatically on new clusters. You can prevent this behavior by configuring the install-config.yaml file before creating the cluster. Disabling network observability prevents the installation of the Network Observability Operator and reduces the cluster’s base resource footprint.
You are enabling the NetworkObservabilityInstall feature gate for Technology Preview.
You have access to the install-config.yaml file before creating installation manifests.
Edit the install-config.yaml file before creating installation manifests.
Add the networkObservability field to the networking section:
apiVersion: v1
baseDomain: example.com
metadata:
name: my-cluster
networking:
networkObservability:
installationPolicy: NoAction
# ...
where:
networking.networkObservability.installationPolicyControls whether the Network Observability Operator is installed during cluster creation. Set to NoAction to prevent installation on all cluster types. Set to InstallAndEnable to force installation. Omit this field to use the default behavior: automatic installation on multi-node and Hosted Control Plane clusters, but not on single-node OpenShift clusters. Valid values are InstallAndEnable, NoAction, or omitted (empty string default).
Configure the FeatureGate manifest in your installation directory to include the NetworkObservabilityInstall feature gate under the designated feature set.
Complete the cluster installation using your platform’s standard deployment method.
Verify that the Network Observability Operator is not installed by running the following command:
$ oc get pods -n netobserv-operator
The output should indicate that the namespace does not exist:
Error from server (NotFound): namespaces "netobserv-operator" not found
Verify that the Network Observability Operator and FlowCollector are running after automatic installation during cluster creation.
You created the cluster with the NetworkObservabilityInstall feature gate enabled.
You have access to the cluster as a user with the cluster-admin role.
The OpenShift CLI (oc) is installed.
Verify that the Network Observability Operator is installed and running by running the following command:
$ oc get pods -n netobserv-operator
NAME READY STATUS RESTARTS AGE
netobserv-controller-manager-6d4b8d9f8c-xk7lm 2/2 Running 0 5m
Verify that the cluster has a FlowCollector resource by running the following command:
$ oc get flowcollector cluster
NAME AGENT SAMPLING DEPLOYMENT MODEL STATUS
cluster eBPF 400 Service Ready
Verify that network traffic data is available in the web console by navigating to Observe → Network Traffic.
Understand how upgrading from OKD 4.22 to 5.0 affects automatic network observability installation when the NetworkObservabilityInstall feature gate is enabled.
When you upgrade an existing cluster from OKD 4.22 to 5.0 with the NetworkObservabilityInstall feature gate enabled, the CNO evaluates the installationPolicy field in the Network custom resource (CR) to determine whether to install network observability.
If the NetworkObservabilityInstall feature gate is not enabled, automatic network observability installation does not occur during the upgrade.
When the feature gate is enabled and you upgrade from OKD 4.22 to 5.0, the following behaviors apply by default:
If the Network CR does not have the networkObservability.installationPolicy field configured, the CNO automatically installs network observability on multi-node and Hosted Control Plane clusters.
Single-node OpenShift clusters do not automatically install network observability due to resource constraints.
If network observability is already installed by using the standard Operator installation method, no changes occur. The automatic installation process detects the existing installation and takes no action.
To prevent automatic installation during the upgrade, see "Prevent automatic network observability installation during upgrades".
You can configure the Network custom resource (CR) to prevent the Cluster Network Operator (CNO) from automatically installing network observability during a cluster upgrade.
|
Since automatic network observability installation is a Technology Preview feature in OKD 5.0, network observability is not installed automatically during upgrades unless you enable the |
You have access to the cluster as a user with the cluster-admin role.
The OpenShift CLI (oc) is installed.
Before upgrading your cluster, configure the Network CR to prevent automatic network observability installation. Choose one of the following methods:
Use oc patch by running the following command:
$ oc patch network.config.openshift.io cluster --type=merge -p '
spec:
networkObservability:
installationPolicy: NoAction
'
Alternatively, edit the Network CR manually by running the following command:
$ oc edit network.config.openshift.io cluster
Add the NoAction installation policy to the spec section:
apiVersion: config.openshift.io/v1
kind: Network
metadata:
name: cluster
spec:
networkObservability:
installationPolicy: NoAction
# ...
Save and close the file to apply the changes.
Proceed with the cluster upgrade. The CNO skips the network observability installation during and after the upgrade process.